MentalBench (DSM Diagnosis) - Incomplete Patient Narratives: leaderboard
Metric: Accuracy (%) on the 3,450 Type 2 cases (colloquial patient self-reports with diagnostic information left incomplete; single answer); cases synthesized from a psychiatrist-built DSM-5 knowledge graph by three LLMs and validated by a psychiatrist and a clinical psychologist; four options per case; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 91.16 | #138 |
| 2 | GPT-5.1 | 90.43 | #131 |
| 3 | GPT-5 Mini | 88.46 | #176 |
| 4 | Gemini 2.5 Pro | 88.2 | #145 |
| 5 | Claude Haiku 4.5 | 87.39 | #271 |
| 6 | Gemini 2.5 Flash | 85.27 | #237 |
| 7 | Qwen 3 235B A22B | 84.17 | #304 |
| 8 | Gemma 3 27B (IT) | 84.09 | #509 |
| 9 | Qwen 3 32B | 83.13 | #424 |
| 10 | Qwen 2.5 72B Instruct | 83.1 | #436 |
| 11 | Qwen 2.5 32B Instruct | 83.04 | #491 |
| 12 | Qwen 2.5 14B Instruct | 82.26 | #634 |
| 13 | Llama 3.1 70B Instruct | 80.81 | #548 |
| 14 | GPT-4o | 79.97 | #333 |
| 15 | Gemma 3 12B (IT) | 78.32 | #655 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=mentalbench-dsm-diagnosis-incomplete-patient-narratives · How It Works · Data refreshed daily, snapshot 2026-10-11.